RFC 4718
IKEv2 Clarifications and Implementation Guidelines
4. Creating CHILD_SAs
4.1. Creating SAs with the CREATE_CHILD_SA Exchange
Section 1.3's organization does not lead to clear understanding of what is needed in which environment. The section can be reorganized with subsections for each use of the CREATE_CHILD_SA exchange (creating child SAs, rekeying IKE SAs, and rekeying child SAs.) The new Section 1.3 with subsections and the above changes might look like the following. NEW-1.3 The CREATE_CHILD_SA Exchange The CREATE_CHILD_SA Exchange is used to create new CHILD_SAs and to rekey both IKE_SAs and CHILD_SAs. This exchange consists of a single request/response pair, and some of its function was referred to as a phase 2 exchange in IKEv1. It MAY be initiated by either end of the IKE_SA after the initial exchanges are completed. All messages following the initial exchange are cryptographically protected using the cryptographic algorithms and keys negotiated in the first two messages of the IKE exchange. These subsequent messages use the syntax of the Encrypted Payload described in section 3.14. All subsequent messages include an Encrypted Payload, even if they are referred to in the text as "empty". The CREATE_CHILD_SA is used for rekeying IKE_SAs and CHILD_SAs. This section describes the first part of rekeying, the creation of new SAs; Section 2.8 covers the mechanics of rekeying, including moving traffic from old to new SAs and the deletion of the old SAs. The two sections must be read together to understand the entire process of rekeying.
Either endpoint may initiate a CREATE_CHILD_SA exchange, so in
this section the term initiator refers to the endpoint
initiating this exchange. An implementation MAY refuse all
CREATE_CHILD_SA requests within an IKE_SA.
The CREATE_CHILD_SA request MAY optionally contain a KE payload
for an additional Diffie-Hellman exchange to enable stronger
guarantees of forward secrecy for the CHILD_SA or IKE_SA. The
keying material for the SA is a function of SK_d established
during the establishment of the IKE_SA, the nonces exchanged
during the CREATE_CHILD_SA exchange, and the Diffie-Hellman
value (if KE payloads are included in the CREATE_CHILD_SA
exchange). The details are described in sections 2.17 and 2.18.
If a CREATE_CHILD_SA exchange includes a KEi payload, at least
one of the SA offers MUST include the Diffie-Hellman group of
the KEi. The Diffie-Hellman group of the KEi MUST be an element
of the group the initiator expects the responder to accept
(additional Diffie-Hellman groups can be proposed). If the
responder rejects the Diffie-Hellman group of the KEi payload,
the responder MUST reject the request and indicate its preferred
Diffie-Hellman group in the INVALID_KE_PAYLOAD Notification
payload. In the case of such a rejection, the CREATE_CHILD_SA
exchange fails, and the initiator SHOULD retry the exchange with
a Diffie-Hellman proposal and KEi in the group that the
responder gave in the INVALID_KE_PAYLOAD.
NEW-1.3.1 Creating New CHILD_SAs with the CREATE_CHILD_SA Exchange
A CHILD_SA may be created by sending a CREATE_CHILD_SA request.
The CREATE_CHILD_SA request for creating a new CHILD_SA is:
Initiator Responder
----------- -----------
HDR, SK {[N+], SA, Ni, [KEi],
TSi, TSr} -->
The initiator sends SA offer(s) in the SA payload, a nonce in
the Ni payload, optionally a Diffie-Hellman value in the KEi
payload, and the proposed traffic selectors for the proposed
CHILD_SA in the TSi and TSr payloads. The request can also
contain Notify payloads that specify additional details for the
CHILD_SA: these include IPCOMP_SUPPORTED, USE_TRANSPORT_MODE,
ESP_TFC_PADDING_NOT_SUPPORTED, and NON_FIRST_FRAGMENTS_ALSO.
The CREATE_CHILD_SA response for creating a new CHILD_SA is:
<-- HDR, SK {[N+], SA, Nr,
[KEr], TSi, TSr}
The responder replies with the accepted offer in an SA payload,
and a Diffie-Hellman value in the KEr payload if KEi was
included in the request and the selected cryptographic suite
includes that group. As with the request, optional Notification
payloads can specify additional details for the CHILD_SA.
The traffic selectors for traffic to be sent on that SA are
specified in the TS payloads in the response, which may be a
subset of what the initiator of the CHILD_SA proposed.
The text about rekeying SAs can be found in Section 5.1 of this
document.
4.2. Creating an IKE_SA without a CHILD_SA
CHILD_SAs can be created either by being piggybacked on the IKE_AUTH
exchange, or using a separate CREATE_CHILD_SA exchange. The
specification is not clear about what happens if creating the
CHILD_SA during the IKE_AUTH exchange fails for some reason.
Our recommendation in this situation is that the IKE_SA is created as
usual. This is also in line with how the CREATE_CHILD_SA exchange
works: a failure to create a CHILD_SA does not close the IKE_SA.
The list of responses in the IKE_AUTH exchange that do not prevent an
IKE_SA from being set up include at least the following:
NO_PROPOSAL_CHOSEN, TS_UNACCEPTABLE, SINGLE_PAIR_REQUIRED,
INTERNAL_ADDRESS_FAILURE, and FAILED_CP_REQUIRED.
(References: "Questions about internal address" thread, April 2005.)
4.3. Diffie-Hellman for First CHILD_SA
Section 1.2 shows that IKE_AUTH messages do not contain KEi/KEr or
Ni/Nr payloads. This implies that the SA payload in IKE_AUTH
exchange cannot contain Transform Type 4 (Diffie-Hellman Group) with
any other value than NONE. Implementations should probably leave the
transform out entirely in this case.
4.4. Extended Sequence Numbers (ESN) Transform
The description of the ESN transform in Section 3.3 has be proved difficult to understand. The ESN transform has the following meaning: o A proposal containing one ESN transform with value 0 means "do not use extended sequence numbers". o A proposal containing one ESN transform with value 1 means "use extended sequence numbers". o A proposal containing two ESN transforms with values 0 and 1 means "I support both normal and extended sequence numbers, you choose". (Obviously this case is only allowed in requests; the response will contain only one ESN transform.) In most cases, the exchange initiator will include either the first or third alternative in its SA payload. The second alternative is rarely useful for the initiator: it means that using normal sequence numbers is not acceptable (so if the responder does not support ESNs, the exchange will fail with NO_PROPOSAL_CHOSEN). Note that including the ESN transform is mandatory when creating ESP/AH SAs (it was optional in earlier drafts of the IKEv2 specification). (References: "Technical change needed to IKEv2 before publication", "STRAW POLL: Dealing with the ESN negotiation interop issue in IKEv2" and "Results of straw poll regarding: IKEv2 interoperability issue" threads, March-April 2005.)4.5. Negotiation of ESP_TFC_PADDING_NOT_SUPPORTED
The description of ESP_TFC_PADDING_NOT_SUPPORTED notification in Section 3.10.1 says that "This notification asserts that the sending endpoint will NOT accept packets that contain Flow Confidentiality (TFC) padding". However, the text does not say in which messages this notification should be included, or whether the scope of this notification is a single CHILD_SA or all CHILD_SAs of the peer. Our interpretation is that the scope is a single CHILD_SA, and thus this notification is included in messages containing an SA payload negotiating a CHILD_SA. If neither endpoint accepts TFC padding, this notification will be included in both the request proposing an SA and the response accepting it. If this notification is included
in only one of the messages, TFC padding can still be sent in one direction.4.6. Negotiation of NON_FIRST_FRAGMENTS_ALSO
NON_FIRST_FRAGMENTS_ALSO notification is described in Section 3.10.1 simply as "Used for fragmentation control. See [RFC4301] for explanation." [RFC4301] says "Implementations that will transmit non-initial fragments on a tunnel mode SA that makes use of non-trivial port (or ICMP type/code or MH type) selectors MUST notify a peer via the IKE NOTIFY NON_FIRST_FRAGMENTS_ALSO payload. The peer MUST reject this proposal if it will not accept non-initial fragments in this context. If an implementation does not successfully negotiate transmission of non-initial fragments for such an SA, it MUST NOT send such fragments over the SA." However, it is not clear exactly how the negotiation works. Our interpretation is that the negotiation works the same way as for IPCOMP_SUPPORTED and USE_TRANSPORT_MODE: sending non-first fragments is enabled only if NON_FIRST_FRAGMENTS_ALSO notification is included in both the request proposing an SA and the response accepting it. In other words, if the peer "rejects this proposal", it only omits NON_FIRST_FRAGMENTS_ALSO notification from the response, but does not reject the whole CHILD_SA creation.4.7. Semantics of Complex Traffic Selector Payloads
As described in Section 3.13, the TSi/TSr payloads can include one or more individual traffic selectors. There is no requirement that TSi and TSr contain the same number of individual traffic selectors. Thus, they are interpreted as follows: a packet matches a given TSi/TSr if it matches at least one of the individual selectors in TSi, and at least one of the individual selectors in TSr. For instance, the following traffic selectors: TSi = ((17, 100, 192.0.1.66-192.0.1.66), (17, 200, 192.0.1.66-192.0.1.66)) TSr = ((17, 300, 0.0.0.0-255.255.255.255), (17, 400, 0.0.0.0-255.255.255.255)) would match UDP packets from 192.0.1.66 to anywhere, with any of the four combinations of source/destination ports (100,300), (100,400), (200,300), and (200, 400).
This implies that some types of policies may require several CHILD_SA pairs. For instance, a policy matching only source/destination ports (100,300) and (200,400), but not the other two combinations, cannot be negotiated as a single CHILD_SA pair using IKEv2. (References: "IKEv2 Traffic Selectors?" thread, Feb 2005.)4.8. ICMP Type/Code in Traffic Selector Payloads
The traffic selector types 7 and 8 can also refer to ICMP type and code fields. As described in Section 3.13.1, "For the ICMP protocol, the two one-octet fields Type and Code are treated as a single 16-bit integer (with Type in the most significant eight bits and Code in the least significant eight bits) port number for the purposes of filtering based on this field." Since ICMP packets do not have separate source and destination port fields, there is some room for confusion what exactly the four TS payloads (two in the request, two in the response, each containing both start and end port fields) should contain. The answer to this question can be found from [RFC4301] Section 4.4.1.3. To give a concrete example, if a host at 192.0.1.234 wants to create a transport mode SA for sending "Destination Unreachable" packets (ICMPv4 type 3) to 192.0.2.155, but is not willing to receive them over this SA pair, the CREATE_CHILD_SA exchange would look like this: Initiator Responder ----------- ----------- HDR, SK { N(USE_TRANSPORT_MODE), SA, Ni, TSi(1, 0x0300-0x03FF, 192.0.1.234-192.0.1.234), TSr(1, 65535-0, 192.0.2.155-192.0.2.155) } --> <-- HDR, SK { N(USE_TRANSPORT_MODE), SA, Nr, TSi(1, 0x0300-0x03FF, 192.0.1.234-192.0.1.234), TSr(1, 65535-0, 192.0.2.155-192.0.2.155) } Since IKEv2 always creates IPsec SAs in pairs, two SAs are also created in this case, even though the second SA is never used for data traffic. An exchange creating an SA pair that can be used both for sending and receiving "Destination Unreachable" places the same value in all the port:
Initiator Responder
----------- -----------
HDR, SK { N(USE_TRANSPORT_MODE), SA, Ni,
TSi(1, 0x0300-0x03FF, 192.0.1.234-192.0.1.234),
TSr(1, 0x0300-0x03FF, 192.0.2.155-192.0.2.155) } -->
<-- HDR, SK { N(USE_TRANSPORT_MODE), SA, Nr,
TSi(1, 0x0300-0x03FF, 192.0.1.234-192.0.1.234),
TSr(1, 0x0300-0x03FF, 192.0.2.155-192.0.2.155) }
(References: "ICMP and MH TSs for IKEv2" thread, Sep 2005.)
4.9. Mobility Header in Traffic Selector Payloads
Traffic selectors can use IP Protocol ID 135 to match the IPv6
mobility header [MIPv6]. However, the IKEv2 specification does not
define how to represent the "MH Type" field in traffic selectors.
At some point, it was expected that this will be defined in a
separate document later. However, [RFC4301] says that "For IKE, the
IPv6 mobility header message type (MH type) is placed in the most
significant eight bits of the 16 bit local "port" selector". The
direction semantics of TSi/TSr port fields are the same as for ICMP
and are described in the previous section.
(References: Tero Kivinen's mail "Issue #86: Add IPv6 mobility header
message type as selector", 2003-10-14. "ICMP and MH TSs for IKEv2"
thread, Sep 2005.)
4.10. Narrowing the Traffic Selectors
Section 2.9 describes how traffic selectors are negotiated when
creating a CHILD_SA. A more concise summary of the narrowing process
is presented below.
o If the responder's policy does not allow any part of the traffic
covered by TSi/TSr, it responds with TS_UNACCEPTABLE.
o If the responder's policy allows the entire set of traffic covered
by TSi/TSr, no narrowing is necessary, and the responder can
return the same TSi/TSr values.
o Otherwise, narrowing is needed. If the responder's policy allows
all traffic covered by TSi[1]/TSr[1] (the first traffic selectors
in TSi/TSr) but not entire TSi/TSr, the responder narrows to an
acceptable subset of TSi/TSr that includes TSi[1]/TSr[1].
o If the responder's policy does not allow all traffic covered by
TSi[1]/TSr[1], but does allow some parts of TSi/TSr, it narrows to
an acceptable subset of TSi/TSr.
In the last two cases, there may be several subsets that are
acceptable (but their union is not); in this case, the responder
arbitrarily chooses one of them and includes ADDITIONAL_TS_POSSIBLE
notification in the response.
4.11. SINGLE_PAIR_REQUIRED
The description of the SINGLE_PAIR_REQUIRED notify payload in
Sections 2.9 and 3.10.1 is not fully consistent.
We do not attempt to describe this payload in this document either,
since it is expected that most implementations will not have policies
that require separate SAs for each address pair.
Thus, if only some part (or parts) of the TSi/TSr proposed by the
initiator is (are) acceptable to the responder, most responders
should simply narrow TSi/TSr to an acceptable subset (as described in
the last two paragraphs of Section 2.9), rather than use
SINGLE_PAIR_REQUIRED.
4.12. Traffic Selectors Violating Own Policy
Section 2.9 describes traffic selector negotiation in great detail.
One aspect of this negotiation that may need some clarification is
that when creating a new SA, the initiator should not propose traffic
selectors that violate its own policy. If this rule is not followed,
valid traffic may be dropped.
This is best illustrated by an example. Suppose that host A has a
policy whose effect is that traffic to 192.0.1.66 is sent via host B
encrypted using Advanced Encryption Standard (AES), and traffic to
all other hosts in 192.0.1.0/24 is also sent via B, but encrypted
using Triple Data Encryption Standard (3DES). Suppose also that host
B accepts any combination of AES and 3DES.
If host A now proposes an SA that uses 3DES, and includes TSr
containing (192.0.1.0-192.0.1.0.255), this will be accepted by host
B. Now, host B can also use this SA to send traffic from 192.0.1.66,
but those packets will be dropped by A since it requires the use of
AES for those traffic. Even if host A creates a new SA only for
192.0.1.66 that uses AES, host B may freely continue to use the first
SA for the traffic. In this situation, when proposing the SA, host A
should have followed its own policy, and included a TSr containing
((192.0.1.0-192.0.1.65),(192.0.1.67-192.0.1.255)) instead.
In general, if (1) the initiator makes a proposal "for traffic X (TSi/TSr), do SA", and (2) for some subset X' of X, the initiator does not actually accept traffic X' with SA, and (3) the initiator would be willing to accept traffic X' with some SA' (!=SA), valid traffic can be unnecessarily dropped since the responder can apply either SA or SA' to traffic X'. (References: "Question about "narrowing" ..." thread, Feb 2005. "IKEv2 needs a "policy usage mode"..." thread, Feb 2005. "IKEv2 Traffic Selectors?" thread, Feb 2005. "IKEv2 traffic selector negotiation examples", 2004-08-08.)4.13. Traffic Selector Authorization
IKEv2 relies on information in the Peer Authorization Database (PAD) when determining what kind of IPsec SAs a peer is allowed to create. This process is described in [RFC4301] Section 4.4.3. When a peer requests the creation of an IPsec SA with some traffic selectors, the PAD must contain "Child SA Authorization Data" linking the identity authenticated by IKEv2 and the addresses permitted for traffic selectors. For example, the PAD might be configured so that authenticated identity "sgw23.example.com" is allowed to create IPsec SAs for 192.0.2.0/24, meaning this security gateway is a valid "representative" for these addresses. Host-to-host IPsec requires similar entries, linking, for example, "fooserver4.example.com" with 192.0.1.66/32, meaning this identity a valid "owner" or "representative" of the address in question. As noted in [RFC4301], "It is necessary to impose these constraints on creation of child SAs to prevent an authenticated peer from spoofing IDs associated with other, legitimate peers." In the example given above, a correct configuration of the PAD prevents sgw23 from creating IPsec SAs with address 192.0.1.66 and prevents fooserver4 from creating IPsec SAs with addresses from 192.0.2.0/24. It is important to note that simply sending IKEv2 packets using some particular address does not imply a permission to create IPsec SAs with that address in the traffic selectors. For example, even if sgw23 would be able to spoof its IP address as 192.0.1.66, it could not create IPsec SAs matching fooserver4's traffic. The IKEv2 specification does not specify how exactly IP address assignment using configuration payloads interacts with the PAD. Our interpretation is that when a security gateway assigns an address
using configuration payloads, it also creates a temporary PAD entry linking the authenticated peer identity and the newly allocated inner address. It has been recognized that configuring the PAD correctly may be difficult in some environments. For instance, if IPsec is used between a pair of hosts whose addresses are allocated dynamically using Dynamic Host Configuration Protocol (DHCP), it is extremely difficult to ensure that the PAD specifies the correct "owner" for each IP address. This would require a mechanism to securely convey address assignments from the DHCP server and link them to identities authenticated using IKEv2. Due to this limitation, some vendors have been known to configure their PADs to allow an authenticated peer to create IPsec SAs with traffic selectors containing the same address that was used for the IKEv2 packets. In environments where IP spoofing is possible (i.e., almost everywhere) this essentially allows any peer to create IPsec SAs with any traffic selectors. This is not an appropriate or secure configuration in most circumstances. See [Aura05] for an extensive discussion about this issue, and the limitations of host-to-host IPsec in general.5. Rekeying and Deleting SAs
5.1. Rekeying SAs with the CREATE_CHILD_SA Exchange
Continued from Section 4.1 of this document. NEW-1.3.2 Rekeying IKE_SAs with the CREATE_CHILD_SA Exchange The CREATE_CHILD_SA request for rekeying an IKE_SA is: Initiator Responder ----------- ----------- HDR, SK {SA, Ni, [KEi]} --> The initiator sends SA offer(s) in the SA payload, a nonce in the Ni payload, and optionally a Diffie-Hellman value in the KEi payload. The CREATE_CHILD_SA response for rekeying an IKE_SA is: <-- HDR, SK {SA, Nr, [KEr]}
The responder replies (using the same Message ID to respond)
with the accepted offer in an SA payload, a nonce in the Nr
payload, and, optionally, a Diffie-Hellman value in the KEr
payload.
The new IKE_SA has its message counters set to 0, regardless of
what they were in the earlier IKE_SA. The window size starts at
1 for any new IKE_SA. The new initiator and responder SPIs are
supplied in the SPI fields of the SA payloads.
NEW-1.3.3 Rekeying CHILD_SAs with the CREATE_CHILD_SA Exchange
The CREATE_CHILD_SA request for rekeying a CHILD_SA is:
Initiator Responder
----------- -----------
HDR, SK {N(REKEY_SA), [N+], SA,
Ni, [KEi], TSi, TSr} -->
The leading Notify payload of type REKEY_SA identifies the
CHILD_SA being rekeyed, and it contains the SPI that the initiator
expects in the headers of inbound packets. In addition, the
initiator sends SA offer(s) in the SA payload, a nonce in the Ni
payload, optionally a Diffie-Hellman value in the KEi payload,
and the proposed traffic selectors in the TSi and TSr payloads.
The request can also contain Notify payloads that specify
additional details for the CHILD_SA.
The CREATE_CHILD_SA response for rekeying a CHILD_SA is:
<-- HDR, SK {[N+], SA, Nr,
[KEr], TSi, TSr}
The responder replies with the accepted offer in an SA payload,
and a Diffie-Hellman value in the KEr payload if KEi was
included in the request and the selected cryptographic suite
includes that group.
The traffic selectors for traffic to be sent on that SA are
specified in the TS payloads in the response, which may be a
subset of what the initiator of the CHILD_SA proposed.
5.2. Rekeying the IKE_SA vs. Reauthentication
Rekeying the IKE_SA and reauthentication are different concepts in
IKEv2. Rekeying the IKE_SA establishes new keys for the IKE_SA and
resets the Message ID counters, but it does not authenticate the
parties again (no AUTH or EAP payloads are involved).
While rekeying the IKE_SA may be important in some environments, reauthentication (the verification that the parties still have access to the long-term credentials) is often more important. IKEv2 does not have any special support for reauthentication. Reauthentication is done by creating a new IKE_SA from scratch (using IKE_SA_INIT/IKE_AUTH exchanges, without any REKEY_SA notify payloads), creating new CHILD_SAs within the new IKE_SA (without REKEY_SA notify payloads), and finally deleting the old IKE_SA (which deletes the old CHILD_SAs as well). This means that reauthentication also establishes new keys for the IKE_SA and CHILD_SAs. Therefore, while rekeying can be performed more often than reauthentication, the situation where "authentication lifetime" is shorter than "key lifetime" does not make sense. While creation of a new IKE_SA can be initiated by either party (initiator or responder in the original IKE_SA), the use of EAP authentication and/or configuration payloads means in practice that reauthentication has to be initiated by the same party as the original IKE_SA. IKEv2 base specification does not allow the responder to request reauthentication in this case; however, this functionality is added in [ReAuth]. (References: "Reauthentication in IKEv2" thread, Oct/Nov 2004.)5.3. SPIs When Rekeying the IKE_SA
Section 2.18 says that "New initiator and responder SPIs are supplied in the SPI fields". This refers to the SPI fields in the Proposal structures inside the Security Association (SA) payloads, not the SPI fields in the IKE header. (References: Tom Stiemerling's mail "Rekey IKE SA", 2005-01-24. Geoffrey Huang's reply, 2005-01-24.)5.4. SPI When Rekeying a CHILD_SA
Section 3.10.1 says that in REKEY_SA notifications, "The SPI field identifies the SA being rekeyed." Since CHILD_SAs always exist in pairs, there are two different SPIs. The SPI placed in the REKEY_SA notification is the SPI the exchange initiator would expect in inbound ESP or AH packets (just as in Delete payloads).
5.5. Changing PRFs When Rekeying the IKE_SA
When rekeying the IKE_SA, Section 2.18 says that "SKEYSEED for the new IKE_SA is computed using SK_d from the existing IKE_SA as follows: SKEYSEED = prf(SK_d (old), [g^ir (new)] | Ni | Nr)" If the old and new IKE_SA selected a different PRF, it is not totally clear which PRF should be used. Since the rekeying exchange belongs to the old IKE_SA, it is the old IKE_SA's PRF that is used. This also follows the principle that the same key (the old SK_d) should not be used with multiple cryptographic algorithms. Note that this may work poorly if the new IKE_SA's PRF has a fixed key size, since the output of the PRF may not be of the correct size. This supports our opinion earlier in the document that the use of PRFs with a fixed key size is a bad idea. (References: "Changing PRFs when rekeying the IKE_SA" thread, June 2005.)5.6. Deleting vs. Closing SAs
The IKEv2 specification talks about "closing" and "deleting" SAs, but it is not always clear what exactly is meant. However, other parts of the specification make it clear that when local state related to a CHILD_SA is removed, the SA must also be actively deleted with a Delete payload. In particular, Section 2.4 says that "If an IKE endpoint chooses to delete CHILD_SAs, it MUST send Delete payloads to the other end notifying it of the deletion". Section 1.4 also explains that "ESP and AH SAs always exist in pairs, with one SA in each direction. When an SA is closed, both members of the pair MUST be closed."5.7. Deleting a CHILD_SA Pair
Section 1.4 describes how to delete SA pairs using the Informational exchange: "To delete an SA, an INFORMATIONAL exchange with one or more delete payloads is sent listing the SPIs (as they would be expected in the headers of inbound packets) of the SAs to be deleted. The recipient MUST close the designated SAs."
The "one or more delete payloads" phrase has caused some confusion. You never send delete payloads for the two sides of an SA in a single message. If you have many SAs to delete at the same time (such as the nested example given in that paragraph), you include delete payloads for the inbound half of each SA in your Informational exchange.5.8. Deleting an IKE_SA
Since IKE_SAs do not exist in pairs, it is not totally clear what the response message should contain when the request deleted the IKE_SA. Since there is no information that needs to be sent to the other side (except that the request was received), an empty Informational response seems like the most logical choice. (References: "Question about delete IKE SA" thread, May 2005.)5.9. Who is the original initiator of IKE_SA
In the IKEv2 document, "initiator" refers to the party who initiated the exchange being described, and "original initiator" refers to the party who initiated the whole IKE_SA. However, there is some potential for confusion because the IKE_SA can be rekeyed by either party. To clear up this confusion, we propose that "original initiator" always refers to the party who initiated the exchange that resulted in the current IKE_SA. In other words, if the "original responder" starts rekeying the IKE_SA, that party becomes the "original initiator" of the new IKE_SA. (References: Paul Hoffman's mail "Original initiator in IKEv2", 2005-04-21.)5.10. Comparing Nonces
Section 2.8 about rekeying says that "If redundant SAs are created though such a collision, the SA created with the lowest of the four nonces used in the two exchanges SHOULD be closed by the endpoint that created it."
Here "lowest" uses an octet-by-octet (lexicographical) comparison (instead of, for instance, comparing the nonces as large integers). In other words, start by comparing the first octet; if they're equal, move to the next octet, and so on. If you reach the end of one nonce, that nonce is the lower one. (References: "IKEv2 rekeying question" thread, July 2005.)5.11. Exchange Collisions
Since IKEv2 exchanges can be initiated by both peers, it is possible that two exchanges affecting the same SA partly overlap. This can lead to a situation where the SA state information is temporarily not synchronized, and a peer can receive a request it cannot process in a normal fashion. Some of these corner cases are discussed in the specification, some are not. Obviously, using a window size greater than one leads to infinitely more complex situations, especially if requests are processed out of order. In this section, we concentrate on problems that can arise even with window size 1. (References: "IKEv2: invalid SPI in DELETE payload" thread, Dec 2005/ Jan 2006. "Problem with exchanges collisions" thread, Dec 2005.)5.11.1. Simultaneous CHILD_SA Close
Probably the simplest case happens if both peers decide to close the same CHILD_SA pair at the same time: Host A Host B -------- -------- send req1: D(SPIa) --> <-- send req2: D(SPIb) --> recv req1 <-- send resp1: () recv resp1 recv req2 send resp2: () --> --> recv resp2 This case is described in Section 1.4 and is handled by omitting the Delete payloads from the response messages.
5.11.2. Simultaneous IKE_SA Close
Both peers can also decide to close the IKE_SA at the same time. The desired end result is obvious; however, in certain cases the final exchanges may not be fully completed. Host A Host B -------- -------- send req1: D() --> <-- send req2: D() --> recv req1 At this point, host B should reply as usual (with empty Informational response), close the IKE_SA, and stop retransmitting req2. This is because once host A receives resp1, it may not be able to reply any longer. The situation is symmetric, so host A should behave the same way. Host A Host B -------- -------- <-- send resp1: () send resp2: () Even if neither resp1 nor resp2 ever arrives, the end result is still correct: the IKE_SA is gone. The same happens if host A never receives req2.5.11.3. Simultaneous CHILD_SA Rekeying
Another case that is described in the specification is simultaneous rekeying. Section 2.8 says "If the two ends have the same lifetime policies, it is possible that both will initiate a rekeying at the same time (which will result in redundant SAs). To reduce the probability of this happening, the timing of rekeying requests SHOULD be jittered (delayed by a random amount of time after the need for rekeying is noticed). This form of rekeying may temporarily result in multiple similar SAs between the same pairs of nodes. When there are two SAs eligible to receive packets, a node MUST accept incoming packets through either SA. If redundant SAs are created though such a collision, the SA created with the lowest of the four nonces used in the two exchanges SHOULD be closed by the endpoint that created it."
However, a better explanation on what impact this has on
implementations is needed. Assume that hosts A and B have an
existing IPsec SA pair with SPIs (SPIa1,SPIb1), and both start
rekeying it at the same time:
Host A Host B
-------- --------
send req1: N(REKEY_SA,SPIa1),
SA(..,SPIa2,..),Ni1,.. -->
<-- send req2: N(REKEY_SA,SPIb1),
SA(..,SPIb2,..),Ni2,..
recv req2 <--
At this point, A knows there is a simultaneous rekeying going on.
However, it cannot yet know which of the exchanges will have the
lowest nonce, so it will just note the situation and respond as
usual.
send resp2: SA(..,SPIa3,..),Nr1,.. -->
--> recv req1
Now B also knows that simultaneous rekeying is going on. Similarly
as host A, it has to respond as usual.
<-- send resp1: SA(..,SPIb3,..),Nr2,..
recv resp1 <--
--> recv resp2
At this point, there are three CHILD_SA pairs between A and B (the
old one and two new ones). A and B can now compare the nonces.
Suppose that the lowest nonce was Nr1 in message resp2; in this case,
B (the sender of req2) deletes the redundant new SA, and A (the node
that initiated the surviving rekeyed SA) deletes the old one.
send req3: D(SPIa1) -->
<-- send req4: D(SPIb2)
--> recv req3
<-- send resp4: D(SPIb1)
recv req4 <--
send resp4: D(SPIa3) -->
The rekeying is now finished.
However, there is a second possible sequence of events that can
happen if some packets are lost in the network, resulting in
retransmissions. The rekeying begins as usual, but A's first packet
(req1) is lost.
Host A Host B
-------- --------
send req1: N(REKEY_SA,SPIa1),
SA(..,SPIa2,..),Ni1,.. --> (lost)
<-- send req2: N(REKEY_SA,SPIb1),
SA(..,SPIb2,..),Ni2,..
recv req2 <--
send resp2: SA(..,SPIa3,..),Nr1,.. -->
--> recv resp2
<-- send req3: D(SPIb1)
recv req3 <--
send resp3: D(SPIa1) -->
--> recv resp3
From B's point of view, the rekeying is now completed, and since it
has not yet received A's req1, it does not even know that these was
simultaneous rekeying. However, A will continue retransmitting the
message, and eventually it will reach B.
resend req1 -->
--> recv req1
What should B do in this point? To B, it looks like A is trying to
rekey an SA that no longer exists; thus failing the request with
something non-fatal such as NO_PROPOSAL_CHOSEN seems like a
reasonable approach.
<-- send resp1: N(NO_PROPOSAL_CHOSEN)
recv resp1 <--
When A receives this error, it already knows there was simultaneous
rekeying, so it can ignore the error message.
5.11.4. Simultaneous IKE_SA Rekeying
Probably the most complex case occurs when both peers try to rekey
the IKE_SA at the same time. Basically, the text in Section 2.8
applies to this case as well; however, it is important to ensure that
the CHILD_SAs are inherited by the right IKE_SA.
The case where both endpoints notice the simultaneous rekeying works
the same way as with CHILD_SAs. After the CREATE_CHILD_SA exchanges,
three IKE_SAs exist between A and B; the one containing the lowest
nonce inherits the CHILD_SAs.
However, there is a twist to the other case where one rekeying
finishes first:
Host A Host B
-------- --------
send req1:
SA(..,SPIa1,..),Ni1,.. -->
<-- send req2: SA(..,SPIb1,..),Ni2,..
--> recv req1
<-- send resp1: SA(..,SPIb2,..),Nr2,..
recv resp1 <--
send req3: D() -->
--> recv req3
At this point, host B sees a request to close the IKE_SA. There's
not much more to do than to reply as usual. However, at this point
host B should stop retransmitting req2, since once host A receives
resp3, it will delete all the state associated with the old IKE_SA
and will not be able to reply to it.
<-- send resp3: ()
5.11.5. Closing and Rekeying a CHILD_SA
A case similar to simultaneous rekeying can occur if one peer decides
to close an SA and the other peer tries to rekey it:
Host A Host B
-------- --------
send req1: D(SPIa) -->
<-- send req2: N(REKEY_SA,SPIb),SA,..
--> recv req1
At this point, host B notices that host A is trying to close an SA
that host B is currently rekeying. Replying as usual is probably the
best choice:
<-- send resp1: D(SPIb)
Depending on in which order req2 and resp1 arrive, host A sees either
a request to rekey an SA that it is currently closing, or a request
to rekey an SA that does not exist. In both cases,
NO_PROPOSAL_CHOSEN is probably fine.
recv req2
recv resp1
send resp2: N(NO_PROPOSAL_CHOSEN) -->
--> recv resp2
5.11.6. Closing a New CHILD_SA
Yet another case occurs when host A creates a CHILD_SA pair, but soon thereafter host B decides to delete it (possible because its policy changed): Host A Host B -------- -------- send req1: [N(REKEY_SA,SPIa1)], SA(..,SPIa2,..),.. --> --> recv req1 (lost) <-- send resp1: SA(..,SPIb2,..),.. <-- send req2: D(SPIb2) recv req2 At this point, host A has not yet received message resp1 (and is retransmitting message req1), so it does not recognize SPIb in message req2. What should host A do? One option would be to reply with an empty Informational response. However, this same reply would also be sent if host A has received resp1, but has already sent a new request to delete the SA that was just created. This would lead to a situation where the peers are no longer in sync about which SAs exist between them. However, host B would eventually notice that the other half of the CHILD_SA pair has not been deleted. Section 1.4 describes this case and notes that "a node SHOULD regard half-closed connections as anomalous and audit their existence should they persist", and continues that "if connection state becomes sufficiently messed up, a node MAY close the IKE_SA". Another solution that has been proposed is to reply with an INVALID_SPI notification that contains SPIb. This would explicitly tell host B that the SA was not deleted, so host B could try deleting it again later. However, this usage is not part of the IKEv2 specification and would not be in line with normal use of the INVALID_SPI notification where the data field contains the SPI the recipient of the notification would put in outbound packets. Yet another solution would be to ignore req2 at this time and wait until we have received resp1. However, this alternative has not been fully analyzed at this time; in general, ignoring valid requests is always a bit dangerous, because both endpoints could do it, leading to a deadlock. This document recommends the first alternative.
5.11.7. Rekeying a New CHILD_SA
Yet another case occurs when a CHILD_SA is rekeyed soon after it has been created: Host A Host B -------- -------- send req1: [N(REKEY_SA,SPIa1)], SA(..,SPIa2,..),.. --> (lost) <-- send resp1: SA(..,SPIb2,..),.. <-- send req2: N(REKEY_SA,SPIb2), SA(..,SPIb3,..),.. recv req2 <-- To host A, this looks like a request to rekey an SA that does not exist. Like in the simultaneous rekeying case, replying with NO_PROPOSAL_CHOSEN is probably reasonable: send resp2: N(NO_PROPOSAL_CHOSEN) --> recv resp15.11.8. Collisions with IKE_SA Rekeying
Another set of cases occurs when one peer starts rekeying the IKE_SA at the same time the other peer starts creating, rekeying, or closing a CHILD_SA. Suppose that host B starts creating a CHILD_SA, and soon after, host A starts rekeying the IKE_SA: Host A Host B -------- -------- <-- send req1: SA,Ni1,TSi,TSr send req2: SA,Ni2,.. --> --> recv req2 What should host B do at this point? Replying as usual would seem like a reasonable choice: <-- send resp2: SA,Ni2,.. recv resp2 <-- send req3: D() --> --> recv req3 Now, a problem arises: If host B now replies normally with an empty Informational response, this will cause host A to delete state associated with the IKE_SA. This means host B should stop retransmitting req1. However, host B cannot know whether or not host A has received req1. If host A did receive it, it will move the
CHILD_SA to the new IKE_SA as usual, and the state information will then be out of sync. It seems this situation is tricky to handle correctly. Our proposal is as follows: if a host receives a request to rekey the IKE_SA when it has CHILD_SAs in "half-open" state (currently being created or rekeyed), it should reply with NO_PROPOSAL_CHOSEN. If a host receives a request to create or rekey a CHILD_SA after it has started rekeying the IKE_SA, it should reply with NO_ADDITIONAL_SAS. The case where CHILD_SAs are being closed is even worse. Our recommendation is that if a host receives a request to rekey the IKE_SA when it has CHILD_SAs in "half-closed" state (currently being closed), it should reply with NO_PROPOSAL_CHOSEN. And if a host receives a request to close a CHILD_SA after it has started rekeying the IKE_SA, it should reply with an empty Informational response. This ensures that at least the other peer will eventually notice that the CHILD_SA is still in "half-closed" state and will start a new IKE_SA from scratch.5.11.9. Closing and Rekeying the IKE_SA
The final case considered in this section occurs if one peer decides to close the IKE_SA while the other peer tries to rekey it. Host A Host B -------- -------- send req1: SA(..,SPIa1,..),Ni1 --> <-- send req2: D() --> recv req1 recv req2 <-- At this point, host B should probably reply with NO_PROPOSAL_CHOSEN, and host A should reply as usual, close the IKE_SA, and stop retransmitting req1. <-- send resp1: N(NO_PROPOSAL_CHOSEN) send resp2: () If host A wants to continue communication with B, it can now start a new IKE_SA.5.11.10. Summary
If a host receives a request to rekey: o a CHILD_SA pair that the host is currently trying to close: reply with NO_PROPOSAL_CHOSEN.
o a CHILD_SA pair that the host is currently rekeying: reply as
usual, but prepare to close redundant SAs later based on the
nonces.
o a CHILD_SA pair that does not exist: reply with
NO_PROPOSAL_CHOSEN.
o the IKE_SA, and the host is currently rekeying the IKE_SA: reply
as usual, but prepare to close redundant SAs and move inherited
CHILD_SAs later based on the nonces.
o the IKE_SA, and the host is currently creating, rekeying, or
closing a CHILD_SA: reply with NO_PROPOSAL_CHOSEN.
o the IKE_SA, and the host is currently trying to close the IKE_SA:
reply with NO_PROPOSAL_CHOSEN.
If a host receives a request to close:
o a CHILD_SA pair that the host is currently trying to close: reply
without Delete payloads.
o a CHILD_SA pair that the host is currently rekeying: reply as
usual, with Delete payload.
o a CHILD_SA pair that does not exist: reply without Delete
payloads.
o the IKE_SA, and the host is currently rekeying the IKE_SA: reply
as usual, and forget about our own rekeying request.
o the IKE_SA, and the host is currently trying to close the IKE_SA:
reply as usual, and forget about our own close request.
If a host receives a request to create or rekey a CHILD_SA when it is
currently rekeying the IKE_SA: reply with NO_ADDITIONAL_SAS.
If a host receives a request to delete a CHILD_SA when it is
currently rekeying the IKE_SA: reply without Delete payloads.
5.12. Diffie-Hellman and Rekeying the IKE_SA
There has been some confusion whether doing a new Diffie-Hellman
exchange is mandatory when the IKE_SA is rekeyed.
It seems that this case is allowed by the IKEv2 specification.
Section 2.18 shows the Diffie-Hellman term (g^ir) in brackets.
Section 3.3.3 does not contradict this when it says that including
the D-H transform is mandatory: although including the transform is mandatory, it can contain the value "NONE". However, having the option to skip the Diffie-Hellman exchange when rekeying the IKE_SA does not add useful functionality to the protocol. The main purpose of rekeying the IKE_SA is to ensure that the compromise of old keying material does not provide information about the current keys, or vice versa. This requires performing the Diffie-Hellman exchange when rekeying. Furthermore, it is likely that this option would have been removed from the protocol as unnecessary complexity had it been discussed earlier. Given this, we recommend that implementations should have a hard- coded policy that requires performing a new Diffie-Hellman exchange when rekeying the IKE_SA. In other words, the initiator should not propose the value "NONE" for the D-H transform, and the responder should not accept such a proposal. This policy also implies that a successful exchange rekeying the IKE_SA always includes the KEi/KEr payloads. (References: "Rekeying IKE_SAs with the CREATE_CHILD_SA exhange" thread, Oct 2005. "Comments of draft-eronen-ipsec-ikev2-clarifications-02.txt" thread, Apr 2005.)
(page 37 continued on part 3)
